DocumentCode :
2650487
Title :
SURF and Spatial Association Correspondence applied in extraction and matching of feature points from MR images of deformed tissues
Author :
Zhang, Xubing ; Hirai, Shinichi ; Zhang, Penglin
Author_Institution :
Dept. of Robot., Ritsumeikan Univ., Kusatsu, Japan
fYear :
2010
fDate :
14-18 Dec. 2010
Firstpage :
448
Lastpage :
453
Abstract :
The extraction and matching of feature points is very important for measuring deformation fields of MR images. Current methods cannot extract and match enough feature points correctly when non-rigid soft biological tissues are deformed in MR images. The authors have therefore used SURF to extract feature points from initial MR images, utilizing every point in deformed MR images as feature points. Subsequently, SURF descriptors and Spatial Association Correspondence (SAC) of neighboring pixels are utilized to match the corresponding feature points of the initial and deformed MR images. Finally, by clustering the differences between deformed points matched by SURF-SAC with the corresponding points calculated by affine transformation, most incorrect match points can be eliminated. Our experimental results show that the proposed method can extract and match more correct corresponding feature point pairs than SURF and SIFT methods.
Keywords :
biomedical MRI; feature extraction; image matching; medical image processing; pattern clustering; MR images; SIFT methods; SURF methods; clustering; deformed tissues; feature points extraction; feature points matching; spatial association correspondence; Approximation methods; Deformable models; Detectors; Feature extraction; Magnetic field measurement; Pixel; Robustness; Deformed; Extraction; Feature point; Matching; SURF; Spatial Association Correspondence;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Robotics and Biomimetics (ROBIO), 2010 IEEE International Conference on
Conference_Location :
Tianjin
Print_ISBN :
978-1-4244-9319-7
Type :
conf
DOI :
10.1109/ROBIO.2010.5723368
Filename :
5723368
Link To Document :
بازگشت